Authors: Marzieh Majd (aDepartment of Biobehavioral Health, The Pennsylvania State University, University Park, PA, USA), Erika F. H. Saunders (bDepartment of Psychiatry, Pennsylvania State College of Medicine, Penn State Milton S. Hershey Medical Center, Hershey, PA, USA), Christopher G. Engeland (aDepartment of Biobehavioral Health, The Pennsylvania State University, University Park, PA, USA; cCenter for Healthy Aging, The Pennsylvania State University, University Park, Pennsylvania, PA, USA; dCollege of Nursing, The Pennsylvania State University, University Park, Pennsylvania, USA)
Categories: Article, depressive symptom, cognitive symptom, neurovegetative symptom, somatic symptom, mood, anhedonia, fatigue, cytokine, C-reactive protein
Source: Frontiers in neuroendocrinology
Authors: Marzieh Majd, Erika F. H. Saunders, Christopher G. Engeland
Patients with depressive disorders show a wide range of clinical manifestations including cognitive and neurovegetative symptoms. Importantly, these symptoms can differ in terms of biological etiology, and deconstructing depression into specific symptoms may provide valuable insight into the underlying neurobiology. Little research has examined inflammation in the context of depressive dimensions. Here we conduct a narrative review of the existing literature (21 studies) to elucidate whether the depression-inflammation link is symptom specific. Overall, there is evidence that an association exists between neurovegetative symptoms of depression and inflammation, independent of cognitive symptoms. The same cannot be said of cognitive symptoms and inflammation. There is also some evidence of gender differences in the directionality of the relationship between depression and inflammation. Potential explanations for these findings, limitations of the existing literature and recommendations for future research design are discussed.
Depression is the leading cause of disability worldwide; over 300 million people today suffer from this common disorder (World Health Organization, 2018). Despite multiple treatment modalities available for major depressive disorder (MDD), less than 40% of patients achieve remission following the initial course of treatment (Rush et al., 2006). An increased understanding of the neurobiological correlates of depressive symptoms would better enable the development of novel therapeutic targets and effective treatments, thus reducing disease burden as well as improving quality of life in patients suffering from depressive disorders.
The clinical presentation of depressive disorders varies greatly across patients (American Psychiatric Association, 2013). The diagnostic evaluation of patients with depressive disorders is based on a cluster of signs and symptoms that are defined by the Diagnostic and Statistical Manual of Mental Disorders, 5^th^ edition (DSM-5). Based on DSM-5 criteria, MDD is characterized by the presence of at least five symptoms for two consecutive weeks. In general, depressive symptoms can be classified into three 1) “psychological symptoms,” a broad concept that often includes core symptoms of depressed mood or anhedonia, as well as feelings of worthlessness, guilt, and suicidal thoughts, 2) “cognitive symptoms,” such as impaired ability to think or concentrate, and 3) “neurovegetative symptoms,” such as sleep problems, fatigue or loss of energy, and changes in appetite. Importantly, these distinct dimensions may reflect different underlying etiology, a possibility which has been supported by evidence linking specific depressive symptoms to different brain circuits. For instance, depressed mood has been associated with altered neuronal activity in the medial prefrontal cortex and appears to be mediated by altered serotonergic and noradrenergic pathways. In contrast, anhedonia and decreased motivation have been linked with reduced dopaminergic activity in the mesocortico-limbic pathway (Nutt, 2008; Stahl et al., 2003). Thus, parsing depression into specific symptoms or dimensions may provide insights into its underlying etiology (Vares et al., 2015).
Strong evidence suggests that inflammation plays a role in the pathophysiology of depressive disorders. Patients with MDD have been found to have increased inflammatory markers in blood such as interleukin (IL)-1β, IL-6, tumor necrosis factor (TNF)-α and C-reactive protein (CRP) (Dowlati et al., 2010; Hiles et al., 2012; Howren et al., 2009); importantly, such increases have been associated with poorer outcomes of antidepressant treatment (Eller et al., 2008; O’Brien et al., 2007). Multiple pathways exist by which peripheral inflammation can signal the brain and lead to major depression, such as the vagus nerve, leaky regions in the blood-brain barrier, and cytokine transport systems (Konsman et al., 2002; Quan and Banks, 2007). Through these pathways, peripheral cytokines can affect the central synthesis and reuptake of monoamine neurotransmitters, including dopamine, serotonin, and norepinephrine (for reviews see Capuron and Miller, 2011; Miller et al., 2013). These findings strengthen the idea that inflammation can contribute to the development and progression of MDD by altering brain chemistry.
Given that interferon (IFN) induces the production of proinflammatory cytokines, IFN therapy has become a well-known model to study inflammation-induced depression. IFN treatment in cancer patients elicited two different patterns of behavioral change across 12 weeks (Capuron et al., 2002). Neurovegetative symptoms (e.g., fatigue, sleep disturbances) developed in a large number of patients by week two of therapy. Conversely, symptoms of depressed mood and cognitive dysfunction appeared later in the course of treatment (i.e., at week 8) and occurred more specifically in patients who met criteria for MDD. Neurovegetative symptoms were less responsive than mood/cognitive symptoms to antidepressant therapy. These differences in symptoms, timing, and response to therapy suggest that different mechanisms mediate the manifestation of depressive dimensions (i.e., mood/cognitive vs. neurovegetative symptoms) in inflammation-induced depression (Capuron et al., 2002).
Although elevated inflammation has been well-documented in both clinical and sub-clinical depression (meta-analyses by Dowlati et al., 2010; Hiles et al., 2012; Howren et al., 2009; Liu et al., 2012), significant variability has been found across studies (Dowlati et al., 2010; Howren et al., 2009). These mixed findings may stem from the fact that depression is a heterogeneous disorder in both its symptomatology and pathophysiology (Slavich and Irwin, 2014). Most studies relating the construct of depression to inflammation have examined depressive disorders only as a broad syndrome. Relatively few studies have examined markers of inflammation in relation to specific symptoms or dimensions of depression, and even fewer studies have examined such associations longitudinally.
Studies examining the association between inflammation and dimensions of depression are varied. They have included different populations (e.g., clinical or community samples), employed different depression rating scales (i.e., self-report or interviewer-rated), used varying dimensional classifications, utilized different statistical approaches, and assessed different inflammatory markers. Therefore, a narrative review of these studies is needed to elucidate whether the depression-inflammation link is symptom specific, to identify sources of inconsistencies across studies, and to highlight gaps in the literature and implications for future research. The primary aims of this review are 1) to summarize the published studies that relate dimensions of depression to basal inflammation in blood, and 2) to address the following a) “Are cognitive and neurovegetative symptoms of depression differentially associated with inflammation?”, and if so, b) “Which dimension most strongly relates to inflammation?”. The current review includes inflammatory biomarkers whose associations with depressive disorders have been verified by meta-analyses (Dowlati et al., 2010; Hiles et al., 2012; Howren et al., 2009; Liu et al., 2012): CRP, IL-6, TNF-α, and cytokine receptors such as soluble IL-6 receptor (sIL-6R) and sTNF-R. Given that an increased level of soluble IL-2 receptor (sIL-2R) in major depression was shown in one meta-analysis, sIL-2R is also included in this review (Liu et al., 2012).
A literature search of the PubMed, Web of Science, and PsychINFO electronic databases was performed of peer-reviewed journals published through October 2018. Combinations of the following key terms were “depress*”, “symptom dimension”, “cognitive”, “somatic”, “neurovegetative”, “inflammation”, “cytokine”, “interleukin”, “IL-6”, “TNF-α”, “cytokine receptor”, and “C-reactive protein”. We screened these terms in titles, keywords and abstracts. The reference lists of relevant studies were screened for additional publications of relevance. A total of 286 studies were identified. Studies were then retained if they met the following four they 1) were conducted in samples of patients with depressive disorders who were physically healthy or in community samples, 2) employed a validated instrument to measure depressive symptoms or diagnose MDD, 3) assessed distinct dimensions of depression (neurovegetative symptoms, cognitive symptoms, or both) in addition to total depression severity, and 4) assessed at least one measure of systemic inflammation (e.g., CRP, IL-6, TNF-α) shown to be reliably associated with depressive symptoms. This approach yielded 30 studies.
Comorbidity may alter the clinical manifestation of depression. Indeed, there is strong evidence that patients with comorbid physical illness (e.g., acute coronary syndrome, metabolic syndrome, cancer) often exhibit more neurovegetative symptoms of depression than affective/cognitive symptoms (Capuron et al., 2008; Carney et al., 2012; Martens et al., 2006; Munoz et al., 2007; Udo et al., 2015). In depression with comorbid physical illness, both the presenting symptoms and the magnitude of accompanied inflammation can be influenced by the underlying physical condition and/or treatment strategy. Hence, studies were excluded when physically-ill patients were the target population (e.g., patients with cancer, autoimmune disorders, myocardial infarction); this review focuses on studies which were conducted in 1) patients with depressive disorders who were otherwise healthy, or 2) community samples. Ultimately, 21 studies were included in this review (Figure 1).
It should be noted that three studies investigated additional inflammatory IL-1Ra, IL18 (Michal et al., 2013), IL-10 (Euteneuer et al., 2017), IFN-γ, IL-2, IL-4, IL-5, IL-10, IL-12, IL-13 (Schmidt et al., 2016). In this review, we do not discuss these cytokines in detail as there is less evidence of their associations with depression; results for these additional cytokines are listed in Table 1.
Most studies that were reviewed measured two broad dimensions of depressive “cognitive” and “neurovegetative”. Usually core depressive symptoms (anhedonia, depressed mood) were classified as cognitive symptoms. However, six studies classified “depressed mood” as a separate dimension from cognitive or neurovegetative dimensions (Bremmer et al., 2008; D. J. Deverts et al., 2010; Köhler-forsberg et al., 2017; Krogh et al., 2014; Niles et al., 2018; White et al., 2017). These six studies are discussed in section 3.2.1; note that we retained the “depressed mood” label to describe these results.
Across studies, the terms “psychological”, “nonsomatic”, and “cognitive” symptoms were used interchangeably. Given that the majority of studies labeled these symptoms as “cognitive”, we label these items as “cognitive” throughout this review. Similarly, the terms “neurovegetative” and “somatic” symptoms were used interchangeably across studies. For the purpose of this review, we use the term “neurovegetative” to describe common physical symptoms observed in depressive disorders; this is consistent with the DSM-5 criteria for major depression and includes fatigue, sleep and appetite. We use the term “somatic” as a broader concept to denote symptoms commonly related to medical illness, such as headache, chest pain, palpitations, dizziness, etc. Detailed information regarding these classifications in each study is provided in the tables.
Psychomotor retardation has both a motor and a cognitive component. As a result, there were inconsistencies in the conceptualization of this construct as a neurovegetative or cognitive symptom across studies. Most of the reviewed studies measured psychomotor disturbances subjectively, using a single item that was classified under the neurovegetative dimension. A few studies employed neuropsychological tasks that assess pure motor function (e.g., motor processing speed), pure cognitive function (e.g., cognitive processing speed) or both (for more details see Buyukduraa, et al., 2011). In this review, the term “psychomotor processing” is only used when psychomotor changes were measured objectively, using tasks that assessed both motor and cognitive processing speeds. Motor processing speed on its own, as measured by the finger-tapping test, is referred to as a neurovegetative symptom, as are psychomotor disturbances that were determined through subjective testing (i.e., by questionnaire).
A number of the reviewed studies reported significant associations between inflammation and depressive symptoms prior to adding covariates (e.g., age, gender) to the analysis (i.e., unadjusted analysis). Such associations were often no longer evident/significant once covariates were added to the analysis (i.e., adjusted analysis). For precision and comparability, only results from adjusted analyses are discussed in text and presented in Tables 1–2. A few studies used the actual dimensions of depression as covariates. Given that this provides information on whether there is an association between one dimension and inflammation over and above other dimensions, both the unadjusted and adjusted results for these particular analyses are presented in text and tables.
Twenty-one studies were included in this review; 14 cross-sectional studies and 7 longitudinal studies. Ten studies included patients diagnosed with DSM criteria (III & IV) [8960 participants with depressive disorders in total] (Bremmer et al., 2008; Chang et al., 2012; Dannehl et al., 2014; Duivis et al., 2013; Elovainio et al., 2009; Euteneuer et al., 2017, 2012; Köhler-forsberg et al., 2017; Krogh et al., 2014; Schmidt et al., 2016). Twelve studies were conducted in community populations [63,886 participants in total]. The most commonly assessed inflammatory biomarker was CRP (measured in 19 studies). The most commonly used questionnaire to assess the dimensions of depressive symptoms was the Beck Depression Inventory (BDI) (included in seven studies). Table 1 (cross-sectional studies) and Table 2 (longitudinal studies) are organized alphabetically by study and provide a comprehensive summary that includes sample characteristics (e.g., inclusion/exclusion criteria), biomarkers, depression scale, definition of each dimension of depression, covariates, and the association between inflammatory markers and depressive symptoms. The results are organized by study type (clinical or community sample) and summarized in Table 3 (cross-sectional studies) and Table 4 (longitudinal studies). For a quick overview of results in text, we direct the reader to summary sections 3.2.1.5, 3.2.2.4, and 3.3.2.
Two studies found an association between inflammatory biomarkers and mood/cognitive symptoms independent of covariates. One study measured CRP and IL-6 levels in 112 outpatients with major depressive disorder (MDD) living in Denmark (67.3% female; mean age 41.6 ± 11.5 years), and 57 healthy controls (65.4% female; mean age 40.3 ± 13.1 years) (Krogh et al., 2014). A battery of neuropsychological tests was administered to measure cognitive function (see Table 1 for more details). No significant associations were observed between CRP and IL-6 levels in relation to depressed mood symptoms, or individual mood items (i.e., depressed mood, guilt feelings, suicide) in either MDD patients or healthy controls. In the entire sample, higher CRP was associated with lower psychomotor speed and poorer executive function (β = 2.42, p = 0.016 and β = −1.54, p = 0.001, respectively), whereas higher IL-6 was associated with higher attention (β = 0.79, p = 0.008). In another study among 231 individuals with MDD (61.5% women, mean age 40.5 ± 11.4 years), CRP levels were measured across nine European countries (Köhler-Forberg et al., 2017). In women, but not in men, a positive association was found between CRP levels and depressed mood symptoms (β = 0.11, p = 0.003), cognitive symptoms (β = 0.13, p = 0.01), and the individual items of interest-activity (β = 0.11, p = 0.02) and suicidality (β = 0.12, p = 0.05).
To examine whether there is a unique association of inflammation with mood/cognitive symptoms, three studies treated neurovegetative symptoms as a covariate. Results revealed that the association between mood/cognitive symptoms and inflammation was no longer significant after adjustment for neurovegetative symptoms. In one study, CRP levels were assessed in 5,909 individuals from a representative sample of the English population (54.9% female, mean age 66 years); of these, 628 (10.6%) were taking antidepressants (White et al., 2017). Both the summed score of depressed mood symptoms and the individual items were examined in relation to quartiles of CRP. Quartiles of CRP values were defined as 0.10-0.79 (reference group), 0.80-1.60, 1.61-3.30, 3.31-157.50. The association between depressed mood symptoms and CRP was significant controlling for age, gender and race/ethnicity (p < 0.001); however, this association was no longer significant after including the sum of neurovegetative symptoms as a covariate. These results were similar among participants who took antidepressants and those who did not. Of the depressed mood symptoms, only the item “feeling depressed” was significantly associated with CRP after controlling for covariates including the sum of the other depressive symptoms (ORs = 1.44, 1.36 and 1.23 for the second, third and fourth quartiles of CRP, respectively, compared with the first, p = 0.04). Similar findings were found in those who did not take antidepressants; only the item “feeling depressed” was associated with CRP independent of these covariates (ORs = 1.41, 1.41 and 1.05, p = 0.02).
Another study analyzed data in 15,071 individuals (50.1% female, mean age 47.5 years) drawn from three cross-sectional studies (Jokela et al., 2016). Authors conducted a meta-analysis on results of these three studies; after controlling for gender, age, and race/ethnicity, all individual depression items were significantly associated with CRP levels (i.e., OR = 1.29 for anhedonia, OR = 1.22 for feeling down, depressed or hopeless, OR = 1.20 for feeling bad about yourself, OR = 1.27 for trouble concentrating on things, and OR = 1.33 for thinking you would be better off dead). When including the sum of other depression items besides the outcome symptom as a covariate, the association between anhedonia and CRP became marginally significant (OR = 1.08; p = 0.06), and all other associations became non-significant.
In a sample of 10,149 adults living in the U.S. (49.2% female, mean age 44.3 years), (Case and Stewart, 2014), the cognitive dimension was positively associated with CRP levels while controlling for demographic factors, diabetes and medication use (β = 0.05, p = 0.001). Stratified-analysis by race/ethnicity revealed that the association between cognitive symptoms and CRP was found in non-Hispanic Whites [n = 4858], but not in other groups (i.e., non-Hispanic Blacks [n = 1978], Mexican Americans [n = 2260], other Hispanics [1053]). After additional adjustment for neurovegetative symptoms, this association was no longer significant.
Two studies reported null results after adjustment for covariates. CRP, IL-6 and TNF-α levels were measured in 2861 individuals (67% female (n=1911), mean age 41.9 ± 13.0 years) living in the Netherlands (Duivis et al., 2013); of these, 2231 had a current or past diagnosis of depressive disorders (i.e., MDD, dysthymia) and/or anxiety (i.e., generalized anxiety disorder, panic disorder, social phobia, agoraphobia) and 630 were controls. After controlling for demographics, health status and antidepressant use, there were no associations between cognitive symptoms and CRP (β = 0.021, p = 0.268), IL-6 (β = 0.028, p = 0.128) or TNF-α (β = 0.02, p = 0.273). No gender differences were found.
In another study, CRP levels were measured in 6005 subjects living in Finland (55% female, mean age 53.0 ± 0.2; mean age 51.5 ± 0.3) (Elovainio et al., 2009). The results of this study were reported separately in men and women. In men, cognitive symptoms were not associated with higher CRP levels after controlling for covariates (see Table 1 for a complete list of covariates); in women, no associations were found in unadjusted or fully adjusted models.
Six studies reported null results in their initial analysis. Note that some of these studies did not specify whether these results were obtained from unadjusted or adjusted analyses. In a sample of 1285 subjects aged 65 and over living in Amsterdam (51% female, mean age 75.4 ± 6.6 years), (Bremmer et al., 2008) the severity of depressed mood did not differ between those with higher versus lower levels of CRP or IL-6 in either the subthreshold depression group (i.e., Center for Epidemiologic Studies Depression Scale [CESD] ≥16 who did not meet DSM criteria for MDD; n = 153) or the major depression group (n = 38).
Another study measured CRP levels in 149 MDD patients living in Taiwan (72% female; mean age 38.8 ± 12.4 years) before and after 6 weeks of antidepressant treatment (i.e., fluoxetine or venlafaxine) (Chang et al., 2012). A battery of neuropsychological tests was administered to measure cognitive function (see Table 1 for more details). No association was observed between baseline CRP and either attention or executive function at baseline, whereas higher baseline CRP was associated with poorer executive function following 6 weeks of antidepressant treatment.
In a study by Dannehl et al. (2014), TNF-α and IL-6 levels were measured in 41 MDD patients (60% female, mean age 33.2 ± 12.52 years), and 45 healthy controls living in Germany (64.7% female, mean age 36.5 ± 13.19 years) at baseline and after 4 weeks. No significant associations were observed between cognitive symptoms and TNF-α or IL-6 either at baseline or over 4 weeks.
Levels of soluble cytokine receptors (sIL-2R, sTNF-R1, sTNF-R2) were measured in 25 MDD patients (76% female, mean age = 31.47 ± 10.11) and 22 healthy controls (63.6% female, mean age = 19.66 ± 7.01) living in Germany (Euteneuer et al., 2012). No significant associations were found between cognitive symptoms and sIL-2R, sTNF-R1 or sTNF-R2. In another study by this group, CRP and IL-6 levels were measured in 98 patients with MDD (49% female, mean age = 37.3 ± 12.2) and 30 healthy controls (50% female, mean age = 37.1 ± 12.2) living in Germany (Euteneuer et al., 2017). No significant associations were found between cognitive symptoms and CRP or IL-6.
Finally, in a population-based sample of 4937 subjects aged 35 to 74 years living in Germany, subjects were categorized into two non-depressed (Patient Health Questionnaire-9 (PHQ-9)<10, n=4580; 48.4% female; mean age = 56.1 years) and depressed (PHQ-9≥10, n=357; 59.4% female; mean age = 53.3 years) (Michal et al., 2013). No significant associations were observed between cognitive symptoms and CRP levels (dichotomized at 3 mg/L).
One study found a negative association between cognitive symptoms and inflammation. In 30 MDD patients (56.6% female, mean age = 37.03 ± 13.09) and 30 age- and sex-matched controls (56.6% female, mean age = 37.03 ± 13.07) in Germany, CRP and TNF-α levels were quantified at baseline and after four weeks of antidepressant therapy (Schmidt et al., 2016). No significant associations were observed between the cognitive symptoms and CRP in either group. Higher cognitive symptoms correlated with lower TNF-α in the depressed group (r = −0.40, p = 0.029), but not in controls. No correlations were found between cognitive symptoms and CRP or TNF-α following four weeks of antidepressant treatment.
CRP: Four cross-sectional studies examined the association between CRP and mood symptoms (Bremmer et al., 2008; Köhler-Forsberg et al., 2017; Krogh et al., 2014; White et al., 2017). Of these, one found that depressed mood symptoms were related to elevated CRP levels (Köhler-Forsberg et al., 2017), and one found that the association between depressed mood symptoms and increased CRP levels was no longer significant after adjustment for neurovegetative symptoms (White et al., 2017). In this latter study, the individual item of feeling depressed was associated with higher inflammation independent of other depressive symptoms (White et al., 2017). Two studies reported null results (Bremmer et al., 2008; Krogh et al., 2014).
Among the ten cross-sectional studies relating cognitive symptoms to CRP, eight reported null results (Case and Stewart, 2014; Chang et al., 2012; Duivis et al., 2013; Elovainio et al., 2009; Euteneuer et al., 2017; Jokela et al., 2016; Michal et al., 2013; Schmidt et al., 2016). It is worth mentioning that of these studies, two found a positive association between cognitive symptoms and CRP prior to adjustment for neurovegetative symptoms (Case and Stewart, 2014; Jokela et al., 2016).
Two studies found a positive association between cognitive symptoms and CRP levels, independent of covariates (Köhler-forsberg et al., 2017; Krogh et al., 2014); neither of these analyses adjusted for neurovegetative symptoms, thus no firm conclusion can be made in these studies regarding the associations of cognitive symptoms with inflammation independent of neurovegetative symptoms. Of the ten studies, only two (Chang et al., 2012; Krogh et al., 2014) investigated objective measures of cognitive function in addition to subjective measures; one found worse cognitive functioning in relation to higher CRP (Krogh et al., 2014).
The two studies that examined depressed mood symptoms in relation to IL-6 levels reported null results (Bremmer et al., 2008; Krogh et al., 2014). Among the four studies relating cognitive symptoms to IL-6, one found a negative association (Krogh et al., 2014), and three reported null results (Dannehl et al., 2014; Duivis et al., 2013; Euteneuer et al., 2017).
Among the three studies relating cognitive symptoms to TNF-α levels, one found a negative association (Schmidt et al., 2016), and two reported null results (Dannehl et al., 2014; Duivis et al., 2013).
Euteneuer et al. (2012) found no significant associations between cognitive symptoms and sIL-2R, sTNF-R1 or sTNF-R2.
Three studies tested for a unique association between neurovegetative symptoms and inflammation independent of mood and cognitive symptoms. Results revealed positive associations. Effect sizes associated with these three studies are presented in Table 5. In a study by White et al. (2017) (previously described in 3.2.1.2), the association between neurovegetative symptoms and quartiles of CRP was significant controlling for all covariates including the sum of depressed mood symptoms (ORs = 0.04, 0.05, 0.15 for the second, third and fourth quartiles of CRP, respectively, compared with the first, p < 0.001). This association was evident among those not taking antidepressants (n = 5281, ORs = 0.04, 0.07, 0.17 for the second, third and fourth quartiles of CRP, respectively, compared with the first, p < 0.001), and was marginally significant in those taking antidepressants (n = 628, ORs = 0.14, 0.03, 0.20 for the second, third and fourth quartiles of CRP, respectively, compared with the first, p = 0.08). The association between each individual neurovegetative item and CRP was significant, controlling for covariates including the sum of other depressive symptoms besides the outcome symptom (for ORs = 1.20, 1.31, and 1.97 for the second, third and fourth quartiles of CRP, respectively, compared with the first, p < 0.001; for restless ORs = 1.08, 1.11, and 1.30, p = 0.03; for low energy/ ORs = 1.28, 1.31, 1.45, p = 0.02). Examining antidepressant use changed these associations. In those who took antidepressants, only the item restless sleep was significantly associated with CRP (ORs of 1.56, 1.11, 1.87, p = 0.04) when controlling for covariates. In those who did not take antidepressants, only the item fatigue was associated with CRP (ORs of 1.25, 1.39, 2.12, p < 0.001) independent of covariates. These findings suggest that antidepressant use moderates the relationship between CRP and neurovegetative symptoms, in particular the symptoms of fatigue and low energy/motivation.
Another study found a positive association between neurovegetative symptoms and CRP levels while controlling for demographic factors, diabetes and medication use (β = 0.08, p < 0.001) (Case and Stewart, 2014). This association remained significant when controlling for cognitive symptoms in the model. Additional adjustment for body mass index (BMI) attenuated the effect size of this association by 44.7%, although the association remained significant (β = 0.05, p = 0.001). Stratified-analysis revealed that the relationship between neurovegetative symptoms and CRP was evident in non-Hispanic Whites, but not in other groups (i.e., non-Hispanic Blacks, Mexican Americans, other Hispanics). These findings suggest that race/ethnicity moderates the link between neurovegetative symptoms and inflammation.
In the third study, the individual items of trouble sleeping or sleeping too much, feeling tired or having little energy, poor appetite or overeating, and moving or speaking slowly or too fast were significantly associated with higher CRP levels (OR = 1.27 for sleep changes, OR = 1.32 for fatigue or lack of energy, OR = 1.32 for changes in appetite, OR = 1.20 for psychomotor disturbance) (Jokela et al., 2016). When including the sum of other depressive symptoms as a covariate, these results remained significant.
The other four studies did not covary for mood/cognitive symptoms. In one study, higher CRP was associated with lower motor processing speed as measured by the finger tapping test; associations with other neurovegetative symptoms were not reported in this study (Chang et al., 2012). Another study examined gender differences and found that neurovegetative symptoms were significantly associated with CRP in men (β = 0.03, p = 0.02), but not in women (using a fully adjusted model) (Elovainio et al., 2009). Two studies measured somatic symptoms in MDD patients. One study found that MDD patients with elevated sIL-2R levels had a higher severity of somatoform symptoms (r = 0.59, p = 0.003); no associations were found with respect to sTNF-R1 or sTNF-R2 (Euteneuer et al., 2012).The other study found no associations between somatic symptoms and TNF-α or IL-6 at baseline or over 4 weeks in MDD patients (Dannehl et al., 2014). However, in women with MDD, higher somatoform symptoms during the last 2 years were predictive of increases in TNF-α over the 4-week period (β = 0.31, P = 0.019). This association was not evident in men (β = −0.01, p = 0.930). It should be noted that in this study the assessment of somatic symptoms was subject to recall bias.
Two studies reported null results after adjustment for covariates. In a study by Duivis et al. (2013), positive associations were found between neurovegetative symptoms and CRP (β = 0.124, p < 0.001), IL-6 (β = 0.088, p < 0.001), and TNF-α (β = 0.07, p < 0.001); however, these associations became non-significant after adding unhealthy lifestyle as a covariate. BMI and smoking status mediated these associations. Another study found no association between higher neurovegetative symptoms and the probability of having CRP levels ≥3 mg/L (fully adjusted model) (Michal et al., 2013).
Four studies reported null results in their initial analysis. Note that some of these studies did not specify whether the results were obtained from unadjusted or adjusted analyses. No associations between neurovegetative symptoms and dichotomized IL-6 and CRP levels were observed among 185 subjects with CESD scores ≥16 (i.e., subthreshold depression and major depression groups) (Bremmer et al., 2008). Two studies found no associations between neurovegetative symptoms and IL-6 or CRP in depressed patients (Euteneuer et al., 2017; Krogh et al., 2014). In addition, another study found no associations between neurovegetative symptoms and CRP levels in the entire sample (MDD patients), or separately in men and women (Köhler-Forberg et al., 2017).
One study found a negative association between neurovegetative symptoms and inflammation. No associations between CRP and neurovegetative symptoms were observed in either MDD patients or healthy controls. Significant negative correlations were evident between TNF-α and neurovegetative symptoms (r = −0.638, p < 0.001) among MDD patients (Schmidt et al., 2016). No associations between CRP and neurovegetative symptoms were observed in either MDD patients or healthy controls. Post-hoc analyses revealed that the individual items of loss of pleasure (r = −0.517), loss of interest (r = −0.633) and concentration (r = −0.506) were negatively correlated with TNF-α. After four weeks of antidepressant therapy, no such correlations were found with CRP or TNF-α.
CRP: Among the twelve studies relating neurovegetative/somatic symptoms to CRP levels, six reported null results (Bremmer et al., 2008; Euteneuer et al,. 2017; Köhler-Forsberg et al., 2017; Krogh et al., 2014; Michal et al. 2013; Schmidt et al., 2016), and five found positive associations between CRP levels and either the summed neurovegetative score or individual items (Case and Stewart, 2014; Duivis et al., 2013; Elovaino et al., 2009; Jokela et al., 2016; White et al., 2017). In addition, one study objectively measured motor speed (by the finger tapping test) and found that higher CRP was associated with lower motor speed (Chang et al., 2012). Case and Stewart (2014) found that neurovegetative symptoms positively related to CRP levels independent of cognitive symptoms. White et al. (2017) and Jokela et al. (2016) reported the same pattern of results; the associations between the neurovegetative symptoms and CRP remained significant after adjustment for other depressive symptoms.
IL-6: Among the five studies relating neurovegetative/somatic symptoms to IL-6, four found null results (Bremmer et al., 2008; Euteneuer et al., 2017; Dannehl et al., 2014; Krogh et al., 2014), and one found a positive association, with BMI and smoking status mediating this effect (Duivis et al., 2013).
TNF-α: Among the three studies relating neurovegetative/somatic symptoms to TNF-α, one found a negative correlation (Schmidt et al., 2016) and one found a positive association; BMI and smoking status mediated this effect (Duivis et al., 2013). Dannehl et al. (2014) found that in women with MDD, higher somatoform symptoms during the last 2 years were predictive of increases in TNF-α over a 4-week period
**Soluble cytokine ** Euteneuer et al. (2012) found that patients with elevated sIL-2R levels had a higher severity of somatoform symptoms. No associations were found with respect to sTNF-R1 or sTNF-R2.
In a study among 2731 healthy participants (mean age=119.0 ± 3.9 months, 97.4% British White) living in England, CRP and IL-6 were assessed at age 9 and depressive symptoms were assessed at age 18 years (mean age 213.6 ± 5.1 months) (Chu et al., 2018). CRP and IL-6 levels were analyzed as both categorical and continuous variables. Baseline CRP levels did not predict increases in neurovegetative or cognitive symptoms after adjustment for covariates. However, structural equation modeling (SEM) revealed that baseline IL-6 was associated with both latent cognitive (coefficient = 0.056, p = 0.016) and neurovegetative (coefficient = 0.059, p = 0.013) dimensions. Each SEM analysis adjusted for the other dimension, indicating that IL-6 was a unique predictor of each dimension (i.e., cognitive, neurovegetative). Baseline IL-6 (age 9) was also associated with diurnal variation in mood (Risk Ratio (RR) = 1.75), concentration difficulties (RR = 1.50), fatigue (RR = 1.31) and sleep disturbances (RR = 1.24) at age 18 years. This suggests there may be an association between inflammation and depressive symptoms across development.
A prospective study examined the association between depressive symptoms and CRP levels in 2544 adults (55% female, mean age 40.2 ± 3.55 years, 42% Black) living in the U.S. (Deverts et al., 2010). Higher baseline scores on either neurovegetative symptoms or depressed mood symptoms predicted higher CRP levels at 5-year follow-up. However, only neurovegetative symptoms (β = 0.08, p <0.008) predicted CRP at follow-up independent of other symptoms. These associations were evident in Black participants, but not White participants, suggesting that race/ethnicity moderates the association between neurovegetative symptoms and inflammation. It should be noted that baseline CRP levels did not predict depressive symptoms at follow-up.
Trajectories of depressive symptoms and their associations with CRP were examined in 1166 adolescents (53.5% female, mean age 16.2 ± 0.6 years) living in the Netherlands over 5 years (Duivis et al. (2015). Cross-sectional findings are shown in Table 2. Three trajectories were identified for cognitive symptoms over 5 persistently low, moderate, or high. Four trajectories were identified for neurovegetative symptoms over 5 “persistently low neurovegetative symptoms”, “decreasing neurovegetative symptoms”, “increasing neurovegetative symptoms”, “persistently high neurovegetative symptoms”. The trajectories of both cognitive symptoms and neurovegetative symptoms over a 5-year period were not associated with subsequent CRP (covarying for demographics and health factors).
The associations of CRP and IL-6 levels with cognitive symptoms of depression were examined in approximately 5978 British white-collar civil servants (30% female, mean age 50 ± 6 years) over an average follow-up of 11.8 years (Gimeno et al., 2009). Cross-sectional findings are shown in Table 2. With respect to prospective analysis, higher baseline levels of CRP (n=3339; β = 0.046, p = 0.004) and IL-6 (n = 3298; β = 0.046, p = 0.005) predicted higher cognitive symptoms of depression in men (for CRP: β = 0.058, p = 0.002, for IL-6: β = 0.054, p=0.006), but not in women. In contrast, baseline cognitive symptoms did not predict CRP or IL-6 at follow-up in the entire sample or in either gender. Neurovegetative/somatic symptoms were not measured in this study.
A large prospective study examined the association between depressive symptoms and CRP levels in 13,775 older adults (59% female, mean age 67 ± 10.3 years) living in the U.S. (Niles et al., 2018). Higher baseline CRP levels predicted higher scores on depressed mood (β=0.03, p<0.001), but were not predictive of neurovegetative symptoms, at 4-year follow-up. Neither of the dimensions at baseline predicted change in CRP. A significant interaction of depression dimension by gender was found. In men, higher neurovegetative symptoms at baseline predicted increases in CRP after 4 years (β =0.04, p =0.006), but not in women. Conversely, in women, higher baseline CRP levels predicted increases in neurovegetative symptoms over time (β =0.03, p =0.03), but not in men.
Another longitudinal study in the U.S. examined 263 community-dwelling adults aged 50 to 70 years (51.7% female, mean age 61.0 ± 4.8 years) at baseline and after 6 years (Stewart et al., 2009). Baseline levels of IL-6 and CRP did not predict increases in either neurovegetative or cognitive symptoms. Neither of the depression dimensions predicted change in CRP. However, neurovegetative symptoms (β =0.15, p =0.03), but not cognitive symptoms (β=0.08, p=0.25), predicted change in IL-6. Gender did not moderate any of these relationships.
The association between CRP levels and both depressive symptoms and cognitive decline were examined in 85-year old individuals (n=456) living in the Netherlands (van den Biggelaar et al., 2007). Cognitive function was measured by the Mini-Mental State Examination (MMSE). Cross-sectional findings are shown in Table 2. Subjects (n=267, 63% female) who did not have depressive symptoms and cognitive dysfunction at baseline (i.e., age 85) were followed prospectively for 5 years. Higher baseline CRP levels significantly predicted increases in depressive symptoms, controlling for cognitive function and comorbidities (F= 17.04, p < 0.001). No significant associations were observed between baseline CRP levels and cognitive function at follow-up.
Among the seven longitudinal studies, three studies found that higher baseline neurovegetative/somatic symptoms predicted increases in inflammatory biomarkers over time (Deverts et al., 2010; Niles et al., 2018; Stewart et al., 2009). Conversely, three studies found that inflammation was positively associated with future depressive symptoms. Specifically, over time, higher baseline IL-6 predicted neurovegetative symptoms (Chu et al., 2018), higher baseline CRP predicted depressed mood (Niles et al., 2018), and both higher baseline CRP and IL-6 predicted cognitive symptoms (Chu et al., 2018; Gimeno et al., 2009). One study found no relationship between CRP and cognitive function (van den Biggelaar, et al., 2007).
Twenty-one studies were included in this review (14 cross-sectional & 7 longitudinal). Overall, there is evidence that an association exists between the neurovegetative symptoms of depression and inflammation beyond potentially confounding factors (including mood and cognitive symptoms). Three reviewed cross-sectional studies (Case and Stewart, 2014; Jokela et al., 2016; White et al., 2017) and two longitudinal studies (Chu et al., 2018; Deverts et al., 2010) found that the association between neurovegetative symptoms and elevated inflammation was robust after simultaneous adjustment for covariates that included other symptoms of depression (i.e., mood/cognitive symptoms), suggesting there might be a unique association between neurovegetative symptoms and inflammation; effect sizes for these associations are presented in Table 5. This finding is consistent with the sickness behavior model, which indicates overlap between the neurovegetative symptoms of depression and inflammation-induced sickness symptoms (e.g., psychomotor slowing, fatigue, anhedonia, changes in sleep and appetite, social withdrawal) (Capuron and Miller, 2004). One particular strength of these five studies is that they had large sample sizes (n = 2544 to 15,071), which provided adequate power to detect the association between dimensions of depression and inflammation.
Conversely, three cross-sectional and one longitudinal study found that the cognitive symptoms of depression were not associated with inflammation independent of neurovegetative symptoms. This might be due to several factors. First, there may be a more robust association between neurovegetative symptoms and inflammation. Neurovegetative symptoms of depression (i.e., sleep disturbance and fatigue) might lead to a more sedentary lifestyle and/or unhealthy diet, consequently resulting in increased obesity and inflammation (Khambaty et al., 2014). Importantly, adipose tissue secretes inflammatory cytokines (Shelton and Miller, 2010). In line with this notion, it has been shown that neurovegetative symptoms were positively associated with insulin resistance independent of cognitive symptoms (Austin et al., 2014; Vrany et al., 2016), and this association was largely explained by CRP and BMI levels. Similarly, a longitudinal study showed that neurovegetative symptoms of depression predicted increases in insulin resistance over 6 years, with BMI accounting for 23% of this association (Khambaty et al., 2014). In this context, one reviewed study found that the relationship between neurovegetative symptoms and inflammation was attenuated after adjustment for BMI (Case and Stewart, 2014); another study found that BMI and smoking mediated the relationship between neurovegetative symptoms and inflammation (Duivis et al., 2013). These findings suggest that heightened BMI (e.g., obesity) is a potential pathway linking neurovegetative symptoms of depression with elevated inflammation. In turn, increased inflammation is associated with both insulin resistance and obesity (Vgontzas et al., 2004), which may further promote inflammation and/or exacerbate depressive symptoms.
Second, there might be causal associations between depressive symptoms; for example, insomnia may precede fatigue, and both symptoms may in turn cause cognitive disturbance (Alhola and Polo-kantola, 2007; Durmer et al., 2005; Fried and Nesse, 2015; Walker, 2008). The involvement of inflammation might strengthen these associations. Indeed, neurovegetative symptoms (e.g., sleep, fatigue) and associated inflammation can act as mediators of cognitive dysfunction in depression. For example, sleep deprivation has been associated with both increased inflammation (cytokines) and impaired cognitive performance (Vgontzas et al., 2004). Three of the reviewed longitudinal studies suggest that inflammation may mediate these associations; each of these studies showed that neurovegetative symptoms predicted higher inflammation over time (Deverts et al., 2010; Niles et al., 2018; Stewart et al., 2009). In addition, elevated inflammation can interfere with sleep and exacerbate fatigue, which in turn may lead to cognitive disturbance. In line with this notion, disruption of the sleep cycle (measured by polysomnography) predicted increased fatigue as well as reduced psychomotor speed following chronic IFN therapy (Raison et al., 2010).
Third, alterations in neurocircuitry implicated in inflammation-associated depression might be particularly relevant to the neurovegetative symptoms of depression. Indeed, altered dopaminergic activity in the basal ganglia due to inflammation has been shown to mediate some depressive symptoms (reviewed by Felger and Miller, 2012), particularly fatigue (Capuron et al., 2007; Haroon et al., 2014), psychomotor retardation (Felger et al., 2016; Haroon et al., 2016), and anhedonia (Felger et al., 2016; Haroon et al., 2016). In the present review, positive associations between inflammation and neurovegetative symptoms of depression were observed in a higher proportion of studies for CRP (6 out of 12 cross-sectional studies) than for any given cytokine (e.g., IL-6: 1 out of 5 studies); this might be explained by the fact that CRP is a more stable measure of inflammation than cytokines, due to its longer half-life and stability across the day. Interestingly, elevated CRP levels in blood have been shown to be an indicator of increased inflammation in the central nervous system (CNS) (Felger et al., 2018). Indeed, elevated CRP in cerebrospinal fluid (CSF) was significantly correlated with plasma levels of CRP (r =0.855), IL-6 (r =0.442), TNF-α (r =0.360), IL-6sr (r =0.328), sTNFR2 (r =0.297) and IL-1ra (r =0.429). Given that a higher proportion of studies reported significant associations between CRP and depression dimensions, and that peripheral CRP levels tend to correlate with central inflammation, CRP may have particular relevance to both depression incidence and symptomatology.
The findings from the current review need to be interpreted in light of limitations. In the majority of studies reviewed, assessment of the underlying constructs of depression was limited to one or two questions, which does not allow for a full representation of each measured construct. A strong example of this is cognition, which requires objective assessment across multiple well-established domains (e.g., working memory, processing speed, attention); this construct cannot be measured adequately by subjective assessment via questionnaire.
According to objective neurocognitive testing, patients with MDD may experience cognitive changes in attention, executive functioning, cognitive processing speed, and visual and verbal memory (Russo et al., 2015). In the majority of studies reviewed, cognitive dysfunction was assessed using only one or two self-reported or observer-rated item(s). Major depression has been linked to impairment in multiple cognitive domains, many of which are not captured by depression questionnaires (e.g., BDI, PHQ-9, CESD-20, Hamilton Depression Rating Scale [HAMD]). In addition, core cognitive elements of depression (e.g., “indecisiveness” or “difficulty concentrating”) were not always included as components of the “cognitive” dimension (Gimeno et al., 2009; Köhler-Forsberg et al., 2017). Hence, multiple facets of cognition are missed by the above scales, and the construct of cognitive impairment is likely underrepresented as a whole. Moreover, five studies classified “indecisiveness” and/or “difficulty concentrating” as neurovegetative symptoms (Bremmer et al., 2008; Deverts et al., 2010; Euteneuer et al., 2017; Schmidt et al., 2016; Stewart et al., 2009) which hinders the interpretation of results. The reason for this discrepancy in classifying depressive symptoms, and suggestions for future studies, are discussed in section 4.6.
Only two studies in this review utilized objective measures of cognition. One study found that poorer executive functioning was associated with higher levels of inflammation in the entire sample (i.e., MDD patients and healthy controls), but not in the MDD group alone (Krogh et al., 2014). Another study found no association between CRP and either attention or executive function in MDD (Chang et al., 2012). Given that each study included a sample size between 100 to 150, larger samples might be needed to reliably detect such associations. Subjective and objective measures of cognitive dysfunction do not always correlate among depressed patients, suggesting they tap into different aspects of cognition (Mohn and Rund, 2016; Russo et al., 2015). To comprehensively measure cognition, both subjective assessment and objective cognitive testing are needed (Lam et al., 2014). Taken together, no strong conclusion can be reached from the reviewed studies regarding the association between cognition and inflammation in depression. A closer examination of the associations between objective measures of cognition (i.e., performance on cognitive tests) and inflammation is needed in MDD to clarify which cognitive domains have underlying inflammatory etiology. Clearly, comprehensive cognitive testing would add valuable information in this field of research.
Multiple symptoms such as psychomotor retardation, fatigue, and sleep are commonly classified under the neurovegetative dimension. In the majority of studies reviewed, these symptoms were amalgamated to create a sum-score, representing total severity for the neurovegetative dimension. Importantly, however, these constructs may have distinct underlying etiologies. Below, these constructs are discussed separately in the context of the reviewed studies, and their associations with inflammation are discussed based on the existing literature.
Most of the reviewed studies classified psychomotor retardation as a neurovegetative symptom, although one study classified this item as a depressed mood symptom (Köhler-Forsberg et al, 2017). Psychomotor retardation has both a motor and a cognitive component, which may explain the inconsistencies in the classification of this construct. There are specific tests that assess motor and/or cognitive components (reviewed by Buyukduraa et al., 2011). For example, the finger-tapping test is a measure of pure motor function and the digit symbol substitution test can assess both motor and cognitive components (Buyukdura et al., 2011). Only two studies in this review utilized objective measures of psychomotor processsing. One found that higher inflammation was associated with lower psychomotor speed in the entire sample (i.e., MDD patients and healthy controls) (Krogh et al., 2014). Another study objectively measured motor processing speed using the finger-tapping test; they found that higher CRP was associated with lower motor speed (Chang et al., 2012). Three studies to date have examined the association between inflammation and psychomotor speed among unmedicated patients with MDD using neurocognitive assessments. One study revealed that higher IL-6 was associated with worse psychomotor performance (Goldsmith et al., 2016). Two studies found that elevated CRP was associated with reduced connectivity between the dorsal striatum and ventromedial prefrontal cortex (vmPFC), as well as elevated glutamate concentrations in the left basal ganglia, which in turn correlated with psychomotor slowing (Felger et al., 2016; Haroon et al., 2016). Slower reaction times and motor speeds have been reported following IFN therapy compared to baseline (e.g., Brydon et al., 2008; Capuron et al., 2001; Majer et al., 2008). Hence, there appears to be a relationship between psychomotor retardation and inflammation in depression. To more fully delineate this association, studies are needed that objectively measure both cognitive and motor components in patients with MDD.
Three of the reviewed cross-sectional studies examined the individual item of fatigue. Two found a positive association with inflammation independent of other depressive symptoms (Jokela et al., 2016; White et al., 2017) and one longitudinal study found that baseline inflammation predicted fatigue 9 years later in a community sample (Chu et al., 2018).
Fatigue is a common symptom in inflammation-induced depression; for example, fatigue is commonly reported following IFN therapy (Capuron et al., 2007, 2005, 2002; Majer et al., 2008; Raison et al., 2014; Trask et al., 2000). Neuroimaging studies have shown alterations in the basal ganglia following chronic exposure to such inflammation. For example, IFN-treated patients have exhibited increases in glucose metabolism and glutamate concentrations in the basal ganglia, both of which correlated with fatigue symptoms (Capuron et al., 2007; Haroon et al., 2014). Fatigue includes both mental (i.e., attention) and physical domains (Chaudhuri and Behan, 2000); importantly, these domains have been linked to distinct underlying neurobiology. Overall, in the context of inflammation-associated depression, little research has been conducted to parse the heterogeneity of fatigue and to investigate the underlying inflammation (Mcmahon et al., 2015). In the reviewed studies, fatigue was assessed with only 1-2 items as part of a depression rating scale (e.g., BDI) which does not adequately assess the above domains. Therefore, future studies are needed that use more specific instruments, such as the multidimensional fatigue inventory (MFI) (Stein et al., 2004), to examine how specific domains within fatigue relate to inflammation. Using such MFI subscales, one study found that a positive association between inflammation and fatigue was driven by reduced motivation in patients with MDD (Felger et al., 2018).
Three cross-sectional studies examined the individual item of sleep. Two found a positive association with inflammation independent of other depressive symptoms (Jokela et al., 2016; White et al., 2017) and one longitudinal study found a positive association between IL-6 at baseline and sleep disturbance (measured by questionnaire) 9 years later in a community sample (Chu et al., 2018).
The association between sleep and inflammation is bidirectional. Both sleep disturbance and long sleep duration (i.e., >8 hours per night) have been linked to elevated inflammation (i.e., CRP & IL-6) (meta-analysis by Irwin et al., 2016). In addition, elevated inflammation can disrupt the sleep cycle. For example, IFN therapy over 12 weeks has been linked to impaired sleep quality (as measured by polysomnography) (Raison et al., 2010), which may further promote inflammation and/or exacerbate sleep-wake disruption. Finally, one study in patients with MDD found that the positive association between total depression score and inflammation is driven by difficulties in sleep initiation (measured by polysomnography) (Motivala et al., 2005). Assessing sleep with both subjective and objective (i.e., polysomnography, actigraphy) measures would help clarify which aspects of sleep (e.g., quality, duration, disturbance) relate to inflammation in the context of depressive disorders.
Most of the reviewed studies classified depressed mood under the cognitive dimension; however, six studies (four cross-sectional and two longitudinal) defined “depressed mood” as a separate dimension (Bremmer et al., 2008; Deverts et al., 2010; Krogh et al., 2014; Köhler-Forsberg et al., 2017; Niles et al., 2018; White et al., 2017). Of the four cross-sectional studies, two reported null findings (Bremmer et al., 2008; Krogh et al., 2014) and two found a positive association between depressed mood symptoms and inflammation (Köhler-Forsberg et al., 2017; White et al., 2017). This discrepancy might be due 1) the larger sample size (n=5909) of White et al. (2017), and/or 2) the conceptualization of psychomotor retardation as “depressed mood” by Köhler-Forsberg et al. (2017). In this latter instance, the observed association between depressed mood and inflammation might have been heavily influenced by this particular item.
Of the two longitudinal studies, one reported null findings (Deverts et al., 2010) and one found that baseline inflammation predicted increases in depressed mood 4 years later (Niles et al., 2018). This latter study had older participants (mean of 67 vs. 40 years) and a larger sample size (13,775 vs. 2544). A different longitudinal study (n=5907), which classified mood symptoms (e.g., hopelessness, worthlessness) as cognitive symptoms, found that baseline inflammation predicted higher cognitive symptoms 11.8 years later (Gimeno et al., 2009). Overall, these longitudinal studies support the notion that inflammation predicts depressed mood over time. In support of this, a meta-analysis of longitudinal studies found that inflammation was associated with subsequent depressive symptomatology as a whole (Valkanova et al., 2013). Together, these findings are consistent with the notion that inflammation contributes to the development of depressed mood and other depressive symptoms, possibly by altering activity within neuropathways (reviewed by Capuron and Miller, 2011; Miller et al., 2013).
Most of the reviewed studies classified loss of interest/pleasure under the cognitive dimension, however, three studies classified this as a neurovegetative symptom (Chu et al., 2018; Schmidt et al., 2016; Stewart et al., 2009). Only one reviewed study reported a marginally positive association between the individual item of anhedonia and inflammation independent of other depressive symptoms (Jokela et al., 2016). Given that altered basal ganglia activity has been associated with anhedonic symptoms and linked to inflammation in the context of MDD (Felger and Miller, 2012), the association between inflammation and anhedonia deserves further attention. Interestingly, Felger et al. (2018) measured inflammation in the CSF of patients with MDD and found a positive correlation between IL-6sr and anhedonia. In another study among patients with MDD, elevated plasma CRP was associated with reduced connectivity between the ventral striatum and vmPFC (Felger et al., 2016), and higher glutamate concentrations in the left basal ganglia (Haroon et al., 2016), both of which correlated with anhedonia. Finally, Jha et al. (2018) found that levels of IL-17, Th1-cytokines (IFN-γ, TNF-α), Th2-cytokines (IL-4, IL-5, IL-9, IL-13), and non-T cell-cytokines (IL-1β, IL-1ra, IL-6, IL-8, macrophage inflammatory protein (MIP)-1α, MIP-1β) were associated with greater severity of anhedonia in MDD patients. Clinical studies in which inflammation is induced provide additional support for these associations (e.g., Capuron et al., 2012; Eisenberger et al., 2010). For example, IFN-treated patients had reduced activity in the ventral striatum in response to rewarding stimuli, which correlated with higher scores on anhedonia (Capuron et al., 2012).
Anhedonia, defined as loss of interest or pleasure, is a multifaceted construct; this includes deficits in consummatory and/or anticipatory pleasure which may be distinct in their underlying neurobiology (reviewed by Treadway and Zald, 2011). “Consummatory pleasure” is defined as the experience of pleasure in response to positive stimuli, whereas “anticipatory pleasure” is the pleasure that people experience at the thought of a future event (Strauss et al., 2011). Most depression rating scales assess anhedonia with 1-4 items, which cannot adequately assess the different facets of anhedonia. There are several self-reported questionnaires (e.g., the Snaith-Hamilton Pleasure Scale, Temporal Experience of Pleasure Scale) and behavioral tasks (the Effort-Expenditure for Rewards Task) specific to anhedonia that address the multidimensional nature of anhedonic symptoms (reviewed by Rizvi et al., 2016; Treadway and Zald, 2011). Future studies are needed which employ these assessment tools and explore the relationship between different facets of anhedonia and inflammation.
Overall, depressive disorders are multidimensional constructs and each construct is multifaceted. Given potential differences in underlying neurobiology, simply relying on the overall severity or intensity of depression (i.e., total depression score) as well as dimension sum-scores (e.g., sum-scores on the neurovegetative dimension) might mask important etiological information. Examining the neurobiological correlates of each construct and multiple domains within each construct would help unravel the complex pathophysiology of depression. This approach could have implications in terms of treatment if such distinctions predicted treatment efficacy. As noted earlier, the association between depression and inflammation may be most relevant to particular symptoms, including psychomotor speed, fatigue, sleep, and anhedonia. Evidence suggests that in patients with treatment-resistant depression, only those with high baseline inflammation (CRP > 5mg/L) benefit from anti-inflammatory therapy (Raison et al., 2013). Examining the effects of anti-inflammatory agents on specific symptoms of depression in this subgroup appears to be an important next step in this research domain (Miller and Raison, 2015).
Only 6 of the 14 cross-sectional studies examined gender effects (Case & Stewart, 2014; Dannehl et al., 2014; Diuivis et al., 2013; Elovainio et al., 2009; Euteneuer et al., 2017; Köhler-Forsberg et al., 2017). Of these, three found no gender effects (Case & Stewart, 2014; Diuivis et al., 2013; Euteneuer et al., 2017) and three yielded mixed 1) somatic symptoms were associated with TNF-α only in women (Dannehl et al., 2014), 2) a positive association between CRP and the neurovegetative dimension was found only in men (Elovainio et al., 2009), and 3) a positive association between CRP and mood/cognitive dimensions occurred only in women (Köhler-Forsberg et al., 2017). Because relatively few of the reviewed studies provided data separately for men and women, concrete conclusions cannot be made regarding the moderating role of gender from these cross-sectional studies.
Three of the seven reviewed longitudinal studies examined gender effects (Gimeno et al., 2009; Niles et al., 2018; Stewart et al., 2009). Of these, one found that gender did not moderate the link between depressive dimensions and inflammation (Steward et al., 2009) and one found that baseline inflammation (IL-6, CRP) predicted cognitive symptoms of depression in men 12 years later (Gimeno et al. 2009). Interestingly, Niles et al. (2018) found that higher neurovegetative symptoms at baseline predicted increases in CRP in men, whereas higher CRP at baseline predicted increases in neurovegetative symptoms in women, over a 4-year period; hence, depression predicted inflammation in men whereas inflammation predicted depression in women. One particular strength of this study is a large sample size (n = 13,775) which provided ample power to detect the association between dimensions of depression and inflammation when split by gender.
The literature offers some support of the findings from this last study. Women appear to be more susceptible than men to developing depression as a result of elevated inflammation (reviewed by Derry et al., 2015). For instance, following in vivo immune challenge, elevated inflammation has been associated with depressed mood in women, but not in men (Eisenberger et al., 2009; Moieni et al., 2015). In contrast, large observational studies have shown depression to have been associated with elevated inflammation in men, but not in women (Ford & Erlinger, 2014; Vogelzangs et al., 2012). Further, greater inflammatory responses to bacterial challenge (ex vivo) have been observed in men reporting more depressive symptoms, but not in women (Majd et al., 2018). Overall, this evidence, combined with the results of Niles et al. (2018), suggest that there may be gender differences in the directionality of the relationship between depression and inflammation; more longitudinal studies are needed to delineate the role of gender in such associations. Most of the reviewed studies considered gender as a covariate; however, simply adjusting for gender does not give information about gender interactions (Dorak and Karpuzoglu, 2012). In future studies, researchers are encouraged to examine the depression-inflammation link separately in men and women, along with examining interactions, to better unmask gender effects in such associations.
Looking across studies, many inconsistencies and limitations were obvious.
Of ten studies that included clinical populations (8960 participants in total), four had a sample size of fewer than 50 (Bremmer, et al., 2008; Euteneuer et al., 2012; Schmidt et al, 2016; Dannehl et al., 2014). In contrast, the majority of studies that were conducted in community samples had a sample size greater than 1000 (63,886 participants in total). Compared to community samples, studies in clinical samples were generally quite small and some were underpowered for accurately detecting the relationship between inflammation and symptom dimensions. Future studies with larger sample sizes are needed to explore these associations in patients with MDD.
One source of methodological inconsistency across studies was the adjustment for covariates. Most studies included age, gender and BMI as covariates in analyses. The following covariates were also included in some 67% of studies adjusted for smoking, 52% for chronic illnesses, 43% for education, 43% for physical exercise, 38% for alcohol intake, 38% for medication use, and 24% for race/ethnicity. In this context, O’Connor et al. (2009) provides empirically based recommendations for including the following factors as covariates when measuring inflammation in age, gender, socioeconomic status, medication use (e.g., statin, aspirin, hypertensive agents, antidepressants), alcohol consumption, dietary factors, fitness level, obesity, race/ethnicity, and menopausal status. Adhering to this type of standardization would be valuable as it would both strengthen research findings and facilitate comparisons across studies.
The reviewed studies used different inclusion/exclusion criteria for recruitment. For example, comorbid illness was excluded in some studies (e.g., Euteneuer et al., 2012) but treated as a covariate in others (e.g., Stewart et al., 2009). In a second example, subjects with CRP levels ≥ 10 mg/L were excluded in eight studies (Bremmer et al., 2008; Case et al., 2014; Deverts et al., 2010; Gimeno et al., 2009; Köhler-Forsberg, et al., 2017; Michal et al., 2014; Niles et al. 2018; Stewart et al., 2009). It is important to note that a marked CRP elevation (i.e., CRP ≥10 mg/L) may be indicative of acute inflammatory conditions such as infection or tissue injury; this differs from low-grade inflammation, which might be seen in conditions such as obesity, cigarette smoking and low alcohol consumption (Kushner, 2015). For studies in which CRP ≥ 10 mg/L is included, researchers are highly recommended to report whether the results change without the inclusion of this subgroup.
Nine studies did not report on antidepressant use, six studies adjusted statistically for antidepressant use, four studies included only unmedicated individuals, one study stratified the sample based on antidepressant use, and one study did not adjust statistically as only eight individuals were taking antidepressants. Based on meta-analysis, certain antidepressants such as SSRIs can reduce levels of inflammatory biomarkers (Hannestad et al., 2011; Hiles et al., 2012); hence their use should be reported in scientific studies. If sample size allows, it is important to examine whether the association between specific symptoms of depression and inflammation differs based on antidepressant use, and whether such associations vary across antidepressant classes.
Twelve studies reported the time of day of blood collection (usually between 7am to 10am). In three studies, blood was drawn between 8am to 2pm. One study did not report the time of blood draw and one simply stated it was obtained in the morning. Blood levels of certain inflammatory cytokines (e.g., IL-1β, IL-6, TNF-α) display diurnal variation (Nilsonne et al., 2016; Vgontzas et al., 2005; Zhou et al., 2010); therefore, the timing of blood sampling should be reported and if collection times are sufficiently discrepant this should be controlled for statistically (e.g., by covariance). Evidence shows this is less important for CRP given its stability across the day (Meier-ewert et al., 2001).
The processing of blood samples varied across studies. Twelve studies used serum samples and seven used plasma to measure inflammatory biomarkers. One study utilized dried blood spot and one did not specify whether plasma or serum was used. The majority of studies did not provide information on the type of collection tubes and sample handling. In terms of anticoagulants used for obtaining plasma, EDTA has provided the most consistent results for IL-1β, IL-6, TNF-α, IFN-α and IFN-γ (Banks, 2000). Note that recommendations for both the type of sample (e.g., serum or plasma) and anticoagulant varies across cytokines and assay techniques (for more details see Banks, 2000). There is evidence that plasma is a more sensitive matrix for detecting changes in certain low-abundance cytokines (Rosenberg-Hasson et al., 2014). Regarding sample handling, Banks (2002) recommends processing blood samples as soon as possible to reduce variability in cytokine levels. Samples should be kept at 4°C until centrifuged. We recommend that researchers report on type of sample (e.g., serum or plasma) and collection tube, and provide information on sample handling (e.g., elapsed time between blood collection and processing) to enable better comparison across studies.
Four studies analyzed inflammatory biomarkers using categorical values (e.g., quartiles). Reasons for this 1. the majority of respondents had IL-6 levels below detection limit, thus IL-6 was dichotomized into low (i.e., non-detectable) and high (i.e., detectable) levels, 2. to reduce the effect of extreme values (e.g., outliers) by treating values as categorical, 3. lack of accepted range in defining low-grade inflammation in children, and 4. skewness of values after log-transformation.
Only two studies examined the moderating role of race/ethnicity in the link between depression dimensions and inflammation. One cross-sectional study found a positive association between neurovegetative symptoms and CRP independent of cognitive symptoms in non-Hispanic Whites, but not in other groups (i.e., non-Hispanic Blacks, Mexican Americans, other Hispanics) (Case and Stewart, 2014). The second study was longitudinal and found that neurovegetative symptoms predicted CRP at 5-year follow-up independent of other symptoms in black participants, but not in white participants (Deverts et al., 2010).
In the former study, participants had a mean PHQ-9 score of 2.7 ± 3.7 (across all race/ethnicity groups) reflecting relatively low depressive symptoms, and only 6.4% of respondents had a PHQ-9 score of 10 or greater representing moderate to severe depression. According to the authors, the link between depressive symptoms and inflammation was weak in this sample and only evident in Non-Hispanic Whites due to the larger sample size (n=4858) compared to Non-Hispanic Blacks (n=1978), Mexican Americans (n=2260), and other Hispanics (n=1053). Regarding the latter study, sample sizes across groups were comparable (white [n=1481] and black [n=1063]); however, black participants reported higher depressive symptoms; only 10% of white participants had CESD-20 scores ≥16 (i.e., high depression) compared to 22% of black participants. This higher severity of depressive symptoms may be one reason that the link between depressive symptoms and inflammation was evident only in black individuals. Future studies are needed to better examine the moderating role of race/ethnicity in the link between depressive symptoms and inflammation in racially diverse MDD patients.
Across studies, different approaches were used to classify depressive symptoms including subjective classification and factor analysis (i.e., statistical approach). As a result, despite many studies using identical questionnaire(s), the same symptoms were often classified into different dimensions. As noted earlier, five studies classified indecisiveness and/or concentration difficulty as neurovegetative symptoms. This classification was based on factor analysis that stemmed from prior work (Dozois et al., 1998; Huang and Chen, 2015; Schmael et al., 2009; Shafer, 2006; Uher et al., 2008; Ward, 2006). Only one of these studies used confirmatory factor analysis to validate the classified symptoms in their sample (Chu et al., 2018). Given that factor analysis outcomes can vary across samples, we recommend basing factor analytic techniques on each current data set (i.e., not relying on results from previous data sets). In instances of low statistical power (i.e., sample size), it would be better to classify depressive symptoms based on theoretical concepts and/or empirical evidence in the context of the research question (for details on sample size in factor analysis see Osborne, 2014). Standardizing the method(s) for classifying symptoms into depressive dimensions would allow for better comparisons across studies. An additional limitation is that depressive dimensions were typically quantified with sum-scores from individual questionnaire items; however, these composite scores represent distinct underlying constructs. As a result, individuals with similar sum-scores may exhibit overt differences in symptomatology; for instance, high sleep and low fatigue might produce an equivalent neurovegetative sum-score to low fatigue and high sleep. The utilization of sum-scores in this manner is vague and can obscure the interpretation of findings; this may be especially relevant when linking such scores to biological markers or underlying etiology.
To better understand the psychopathology of mental disorders, recent emphasis has been placed on disaggregating multidimensional constructs (e.g., depressive dimensions) into more homogenous constructs (Fried and Nesse, 2015; Smith et al., 2009). The Research Domain Criteria (RDoC) framework launched by the National Institute of Mental Health has proposed new approaches of classifying mental disorders based on neurobehavioral systems (Cuthbert, 2014). This approach focuses on proximal neurobiological mechanisms rather than symptom-driven classification (Etkin and Cuthbert, 2014). The proposed domains of functioning negative valence systems, positive valence systems, cognitive symptoms, social processes, arousal and regulatory systems, and sensorimotor systems. Each domain entails various constructs and each construct is defined across multiple units of analysis (e.g., genes, circuits, behavior, physiology) (Cuthbert, 2014). Consistent with this notion, the premise of the present review suggests the potential value in clarifying and separating the constructs of depression. In this context, we posit that clustering depressive symptoms on the basis of underlying neurocircuitry (i.e., biological signature) may have merit. Indeed, there is strong evidence that inflammation can alter dopamine function in the basal ganglia, resulting in psychomotor retardation, anhedonia, and fatigue (reviewed by Felger & Miller, 2012). Clustering these symptoms by known associated brain regions (i.e., basal ganglia) may better inform how these symptoms interrelate and also inform treatment.
Taken together, the findings from these 21 studies suggest that an association exists between neurovegetative symptoms and inflammation beyond potentially confounders (including mood and cognitive symptoms). Further, neurovegetative symptoms appear to predict higher inflammation over time. Conversely, cognitive symptoms of depression were not associated with inflammation independent of neurovegetative symptoms. A possible explanation is that neurovegetative symptoms (e.g., sleep, fatigue) may act as mediators of cognitive dysfunction in depression. These findings need to be interpreted in light of the many inconsistencies that existed across studies. The classification of depressive symptoms as “cognitive” or “neurovegetative” often differed between the reviewed studies; a standardized method for classifying dimensions would allow for better comparisons and conclusions to be drawn across future studies. Given that a single scale is typically comprised of only a few items, multiple assessment tools are needed for comprehensive examination of different aspects of depressive symptoms. See Summary Box for suggestions for future research design.
Examining the biological correlates of each construct within depression may provide valuable information regarding the neurobiology of depressive symptoms, beyond what can be learned from focusing on total depression severity. A precise explication and assessment of the various constructs within depression would better capture the heterogeneity of depressive symptomatology and may inform its underlying etiology. Ultimately, such information might help guide clinicians to select medications that target individual symptoms and allow for more exact treatment of depression based on its underlying pathology within a given patient.